Skip to content
Monitoring

Govern AI agents on Datadog

Connect Datadog monitoring to Cordum-governed agents. Trigger agent workflows from Datadog alerts while enforcing governance on every automated response. The Datadog pack bridges observability and autonomous remediation with safety controls.

What this pack does

  • Webhook-triggered agent workflows from Datadog alerts
  • Alert severity-based policy routing
  • Governed auto-remediation pipelines
  • Correlation of Datadog metrics with agent actions

Use cases

Trigger governed remediation when CPU exceeds threshold

Route P1 alerts to approval-gated agent workflows

Log agent responses alongside Datadog incidents

Quick setup

  1. 1Install the Datadog pack: cordumctl pack install datadog
  2. 2Configure Datadog webhook to point to Cordum
  3. 3Map alert severities to governance tiers
  4. 4Enable the pack and configure alert routing

Frequently asked questions

How does a Datadog alert kick off a governed agent workflow?

You point a Datadog webhook at Cordum, and incoming alerts trigger an agent workflow. Before any remediation runs, the action is evaluated by the Safety Kernel, so the path from alert to automated response is governed end to end.

Can different alert severities follow different governance rules?

Yes. Map Datadog severities to governance tiers so that, for example, a P1 alert routes to an approval-gated workflow while a low-severity alert can auto-remediate within constraints. Routing is driven by the severity you configure on each alert.

Can agents mute or change Datadog monitors?

The pack supports read actions like querying metrics and logs by default, and treats monitor changes such as mute and unmute as write actions subject to policy. That keeps an agent from silencing a monitor without an Allow or Require Approval decision.

How do we correlate agent responses with Datadog incidents?

Cordum records each agent action and its governance decision in the audit trail alongside the triggering alert. You can line up Datadog metrics and incidents with the exact remediation an agent took and whether a human approved it.

Ready to govern Datadog?

Other integrations